Perl's rand() function with localtime as seed - perl

Question

Please look at the following piece of code:
srand(localtime);
for (my $ik = 0; $ik < 3; $ik += 1)
{
print int(rand(10)),"\n";
sleep(1);
}
I invoke the above piece of code multiple times with sufficient time(5-10 sec) in between, still the output sequence is the same.
As I have set the seed to localtime every invocation must use a different seed and perhaps generate a different sequence of three numbers, because of the time gap. Why do I get the same sequence again and again.
Note: The code is NOT in a loop, it is in a Perl file which is being executed multiple times.
The documentation says that this seed fails if multiple instances run in the same 'second' leading to the same seed - here it is not the case.
EDIT:: The solution by #simbabque does help but the randomness expected is not gained. Look at my comment on the aforementioned solution below.

Answer 1

Try running this with use strict and use warnings. It will give you:
Argument "Thu Jun 21 13:04:41 2012" isn't numeric in srand at ...
And right there is your problem. localtime returns a string in scalar context. Try using time instead, which returns the unix timestamp as an integer. srand needs a numerical value to work.
If you add a Data::Dumper to it you'll see that the seed with your code is always 1.
no strict; no warnings;
use Data::Dumper;
print Dumper srand(localtime);
for (my $ik = 0; $ik < 3; $ik += 1)
{
print int(rand(10)),"\n";
sleep(1);
}
Says:
$VAR1 = 1;
0
2
6
What you need is:
use strict; use warnings;
srand(time);
for (my $ik = 0; $ik < 3; $ik += 1)
{
print int(rand(10)),"\n";
sleep(1);
}
Edit:
This still is not a very good idea if you want good randomness. The doc says:
In versions of Perl prior to 5.004 the default seed was just the
current time. This isn't a particularly good seed, so many old
programs supply their own seed value (often time ^ $$ or time ^ ($$ +
($$ << 15)) ), but that isn't necessary any more.
I suggest you just omit the call to srand at all unless you actually want reproducable results (i.e. for testing).

Answer 2

In general, there is no reason to expect better randomness by repeatedly seeding a PRNG.
You can use the following script to check what's going on with your original question:
#!/usr/bin/env perl
use strict; use warnings;
use 5.014;
for (1 .. 3) {
my $seq = newseq(3, 5);
printf "Seed = %s\n", $seq->{seed};
my $it = $seq->{generator};
while (defined(my $r = $it->())) {
print "$r\n";
}
sleep 5;
}
sub newseq {
my ($length, $limit) = #_;
$length //= 10;
$limit //= 10;
my $seed = srand(time);
return {
seed => $seed,
generator => sub {
return unless $length-- > 0;
return rand($limit);
},
};
}
However, if you do need statistically independent generators, you can use Math::Random::MT::Auto and create individual PRNG objects:
#!/usr/bin/env perl
use strict; use warnings;
use 5.014;
use strict;
use warnings;
use Math::Random::MT::Auto qw(:!auto);
my $prng1 = Math::Random::MT::Auto->new(SOURCE => '/dev/random');
my $prng2 = Math::Random::MT::Auto->new(SOURCE => 'random_org');
say $prng1->rand();
say $prng2->irand();